pax.c revision 1.14 1 /* $NetBSD: pax.c,v 1.14 2001/10/25 05:33:33 lukem Exp $ */
2
3 /*-
4 * Copyright (c) 1992 Keith Muller.
5 * Copyright (c) 1992, 1993
6 * The Regents of the University of California. All rights reserved.
7 *
8 * This code is derived from software contributed to Berkeley by
9 * Keith Muller of the University of California, San Diego.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. All advertising materials mentioning features or use of this software
20 * must display the following acknowledgement:
21 * This product includes software developed by the University of
22 * California, Berkeley and its contributors.
23 * 4. Neither the name of the University nor the names of its contributors
24 * may be used to endorse or promote products derived from this software
25 * without specific prior written permission.
26 *
27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * SUCH DAMAGE.
38 */
39
40 #include <sys/cdefs.h>
41 #ifndef lint
42 __COPYRIGHT("@(#) Copyright (c) 1992, 1993\n\
43 The Regents of the University of California. All rights reserved.\n");
44 #endif /* not lint */
45
46 #ifndef lint
47 #if 0
48 static char sccsid[] = "@(#)pax.c 8.2 (Berkeley) 4/18/94";
49 #else
50 __RCSID("$NetBSD: pax.c,v 1.14 2001/10/25 05:33:33 lukem Exp $");
51 #endif
52 #endif /* not lint */
53
54 #include <stdio.h>
55 #include <sys/types.h>
56 #include <sys/param.h>
57 #include <sys/stat.h>
58 #include <sys/time.h>
59 #include <sys/resource.h>
60 #include <signal.h>
61 #include <unistd.h>
62 #include <stdlib.h>
63 #include <errno.h>
64 #include "pax.h"
65 #include "extern.h"
66 static int gen_init(void);
67
68 /*
69 * PAX main routines, general globals and some simple start up routines
70 */
71
72 /*
73 * Variables that can be accessed by any routine within pax
74 */
75 int act = DEFOP; /* read/write/append/copy */
76 FSUB *frmt = NULL; /* archive format type */
77 int cflag; /* match all EXCEPT pattern/file */
78 int dflag; /* directory member match only */
79 int iflag; /* interactive file/archive rename */
80 int kflag; /* do not overwrite existing files */
81 int lflag; /* use hard links when possible */
82 int nflag; /* select first archive member match */
83 int tflag; /* restore access time after read */
84 int uflag; /* ignore older modification time files */
85 int vflag; /* produce verbose output */
86 int zflag; /* use gzip */
87 int Aflag; /* honor absolute path */
88 int Dflag; /* same as uflag except inode change time */
89 int Hflag; /* follow command line symlinks (write only) */
90 int Lflag; /* follow symlinks when writing */
91 int Xflag; /* archive files with same device id only */
92 int Yflag; /* same as Dflg except after name mode */
93 int Zflag; /* same as uflg except after name mode */
94 int vfpart; /* is partial verbose output in progress */
95 int patime = 1; /* preserve file access time */
96 int pmtime = 1; /* preserve file modification times */
97 int pfflags = 1; /* preserve file flags */
98 int pmode; /* preserve file mode bits */
99 int pids; /* preserve file uid/gid */
100 int exit_val; /* exit value */
101 int docrc; /* check/create file crc */
102 char *dirptr; /* destination dir in a copy */
103 char *argv0; /* root of argv[0] */
104 sigset_t s_mask; /* signal mask for cleanup critical sect */
105
106 /*
107 * PAX - Portable Archive Interchange
108 *
109 * A utility to read, write, and write lists of the members of archive
110 * files and copy directory hierarchies. A variety of archive formats
111 * are supported (some are described in POSIX 1003.1 10.1):
112 *
113 * ustar - 10.1.1 extended tar interchange format
114 * cpio - 10.1.2 extended cpio interchange format
115 * tar - old BSD 4.3 tar format
116 * binary cpio - old cpio with binary header format
117 * sysVR4 cpio - with and without CRC
118 *
119 * This version is a superset of IEEE Std 1003.2b-d3
120 *
121 * Summary of Extensions to the IEEE Standard:
122 *
123 * 1 READ ENHANCEMENTS
124 * 1.1 Operations which read archives will continue to operate even when
125 * processing archives which may be damaged, truncated, or fail to meet
126 * format specs in several different ways. Damaged sections of archives
127 * are detected and avoided if possible. Attempts will be made to resync
128 * archive read operations even with badly damaged media.
129 * 1.2 Blocksize requirements are not strictly enforced on archive read.
130 * Tapes which have variable sized records can be read without errors.
131 * 1.3 The user can specify via the non-standard option flag -E if error
132 * resync operation should stop on a media error, try a specified number
133 * of times to correct, or try to correct forever.
134 * 1.4 Sparse files (lseek holes) stored on the archive (but stored with blocks
135 * of all zeros will be restored with holes appropriate for the target
136 * filesystem
137 * 1.5 The user is notified whenever something is found during archive
138 * read operations which violates spec (but the read will continue).
139 * 1.6 Multiple archive volumes can be read and may span over different
140 * archive devices
141 * 1.7 Rigidly restores all file attributes exactly as they are stored on the
142 * archive.
143 * 1.8 Modification change time ranges can be specified via multiple -T
144 * options. These allow a user to select files whose modification time
145 * lies within a specific time range.
146 * 1.9 Files can be selected based on owner (user name or uid) via one or more
147 * -U options.
148 * 1.10 Files can be selected based on group (group name or gid) via one o
149 * more -G options.
150 * 1.11 File modification time can be checked against existing file after
151 * name modification (-Z)
152 *
153 * 2 WRITE ENHANCEMENTS
154 * 2.1 Write operation will stop instead of allowing a user to create a flawed
155 * flawed archive (due to any problem).
156 * 2.2 Archives written by pax are forced to strictly conform to both the
157 * archive and pax the spceific format specifications.
158 * 2.3 Blocking size and format is rigidly enforced on writes.
159 * 2.4 Formats which may exhibit header overflow problems (they have fields
160 * too small for large file systems, such as inode number storage), use
161 * routines designed to repair this problem. These techniques still
162 * conform to both pax and format specifications, but no longer truncate
163 * these fields. This removes any restrictions on using these archive
164 * formats on large file systems.
165 * 2.5 Multiple archive volumes can be written and may span over different
166 * archive devices
167 * 2.6 A archive volume record limit allows the user to specify the number
168 * of bytes stored on an archive volume. When reached the user is
169 * prompted for the next archive volume. This is specified with the
170 * non-standard -B flag. The limit is rounded up to the next blocksize.
171 * 2.7 All archive padding during write use zero filled sections. This makes
172 * it much easier to pull data out of flawed archive during read
173 * operations.
174 * 2.8 Access time reset with the -t applies to all file nodes (including
175 * directories).
176 * 2.9 Symbolic links can be followed with -L (optional in the spec).
177 * 2.10 Modification or inode change time ranges can be specified via
178 * multiple -T options. These allow a user to select files whose
179 * modification or inode change time lies within a specific time range.
180 * 2.11 Files can be selected based on owner (user name or uid) via one or more
181 * -U options.
182 * 2.12 Files can be selected based on group (group name or gid) via one o
183 * more -G options.
184 * 2.13 Symlinks which appear on the command line can be followed (without
185 * following other symlinks; -H flag)
186 *
187 * 3 COPY ENHANCEMENTS
188 * 3.1 Sparse files (lseek holes) can be copied without expanding the holes
189 * into zero filled blocks. The file copy is created with holes which are
190 * appropriate for the target filesystem
191 * 3.2 Access time as well as modification time on copied file trees can be
192 * preserved with the appropriate -p options.
193 * 3.3 Access time reset with the -t applies to all file nodes (including
194 * directories).
195 * 3.4 Symbolic links can be followed with -L (optional in the spec).
196 * 3.5 Modification or inode change time ranges can be specified via
197 * multiple -T options. These allow a user to select files whose
198 * modification or inode change time lies within a specific time range.
199 * 3.6 Files can be selected based on owner (user name or uid) via one or more
200 * -U options.
201 * 3.7 Files can be selected based on group (group name or gid) via one o
202 * more -G options.
203 * 3.8 Symlinks which appear on the command line can be followed (without
204 * following other symlinks; -H flag)
205 * 3.9 File inode change time can be checked against existing file before
206 * name modification (-D)
207 * 3.10 File inode change time can be checked against existing file after
208 * name modification (-Y)
209 * 3.11 File modification time can be checked against existing file after
210 * name modification (-Z)
211 *
212 * 4 GENERAL ENHANCEMENTS
213 * 4.1 Internal structure is designed to isolate format dependent and
214 * independent functions. Formats are selected via a format driver table.
215 * This encourages the addition of new archive formats by only having to
216 * write those routines which id, read and write the archive header.
217 */
218
219 /*
220 * main()
221 * parse options, set up and operate as specified by the user.
222 * any operational flaw will set exit_val to non-zero
223 * Return: 0 if ok, 1 otherwise
224 */
225
226 int
227 main(int argc, char **argv)
228 {
229 /*
230 * parse options, determine operational mode, general init
231 */
232 options(argc, argv);
233 if ((gen_init() < 0) || (tty_init() < 0))
234 return(exit_val);
235
236 (void)time(&starttime);
237 #ifdef SIGINFO
238 (void)signal(SIGINFO, ar_summary);
239 #endif
240 /*
241 * select a primary operation mode
242 */
243 switch(act) {
244 case EXTRACT:
245 extract();
246 break;
247 case ARCHIVE:
248 archive();
249 break;
250 case APPND:
251 append();
252 break;
253 case COPY:
254 copy();
255 break;
256 default:
257 case LIST:
258 list();
259 break;
260 }
261 return(exit_val);
262 }
263
264 /*
265 * sig_cleanup()
266 * when interrupted we try to do whatever delayed processing we can.
267 * This is not critical, but we really ought to limit our damage when we
268 * are aborted by the user.
269 * Return:
270 * never....
271 */
272
273 void
274 sig_cleanup(int which_sig)
275 {
276 /*
277 * restore modes and times for any dirs we may have created
278 * or any dirs we may have read. Set vflag and vfpart so the user
279 * will clearly see the message on a line by itself.
280 */
281 vflag = vfpart = 1;
282 if (which_sig == SIGXCPU)
283 tty_warn(0, "Cpu time limit reached, cleaning up.");
284 else
285 tty_warn(0, "Signal caught, cleaning up.");
286
287 ar_close();
288 proc_dir();
289 if (tflag)
290 atdir_end();
291 exit(1);
292 /* NOTREACHED */
293 }
294
295 /*
296 * gen_init()
297 * general setup routines. Not all are required, but they really help
298 * when dealing with a medium to large sized archives.
299 */
300
301 static int
302 gen_init(void)
303 {
304 struct rlimit reslimit;
305 struct sigaction n_hand;
306 struct sigaction o_hand;
307
308 /*
309 * Really needed to handle large archives. We can run out of memory for
310 * internal tables really fast when we have a whole lot of files...
311 */
312 if (getrlimit(RLIMIT_DATA , &reslimit) == 0){
313 reslimit.rlim_cur = reslimit.rlim_max;
314 (void)setrlimit(RLIMIT_DATA , &reslimit);
315 }
316
317 /*
318 * should file size limits be waived? if the os limits us, this is
319 * needed if we want to write a large archive
320 */
321 if (getrlimit(RLIMIT_FSIZE , &reslimit) == 0){
322 reslimit.rlim_cur = reslimit.rlim_max;
323 (void)setrlimit(RLIMIT_FSIZE , &reslimit);
324 }
325
326 /*
327 * increase the size the stack can grow to
328 */
329 if (getrlimit(RLIMIT_STACK , &reslimit) == 0){
330 reslimit.rlim_cur = reslimit.rlim_max;
331 (void)setrlimit(RLIMIT_STACK , &reslimit);
332 }
333
334 /*
335 * not really needed, but doesn't hurt
336 */
337 if (getrlimit(RLIMIT_RSS , &reslimit) == 0){
338 reslimit.rlim_cur = reslimit.rlim_max;
339 (void)setrlimit(RLIMIT_RSS , &reslimit);
340 }
341
342 /*
343 * signal handling to reset stored directory times and modes. Since
344 * we deal with broken pipes via failed writes we ignore it. We also
345 * deal with any file size limit thorugh failed writes. Cpu time
346 * limits are caught and a cleanup is forced.
347 */
348 if ((sigemptyset(&s_mask) < 0) || (sigaddset(&s_mask, SIGTERM) < 0) ||
349 (sigaddset(&s_mask,SIGINT) < 0)||(sigaddset(&s_mask,SIGHUP) < 0) ||
350 (sigaddset(&s_mask,SIGPIPE) < 0)||(sigaddset(&s_mask,SIGQUIT)<0) ||
351 (sigaddset(&s_mask,SIGXCPU) < 0)||(sigaddset(&s_mask,SIGXFSZ)<0)) {
352 tty_warn(1, "Unable to set up signal mask");
353 return(-1);
354 }
355 n_hand.sa_mask = s_mask;
356 n_hand.sa_flags = 0;
357 n_hand.sa_handler = sig_cleanup;
358
359 if ((sigaction(SIGHUP, &n_hand, &o_hand) < 0) &&
360 (o_hand.sa_handler == SIG_IGN) &&
361 (sigaction(SIGHUP, &o_hand, &o_hand) < 0))
362 goto out;
363
364 if ((sigaction(SIGTERM, &n_hand, &o_hand) < 0) &&
365 (o_hand.sa_handler == SIG_IGN) &&
366 (sigaction(SIGTERM, &o_hand, &o_hand) < 0))
367 goto out;
368
369 if ((sigaction(SIGINT, &n_hand, &o_hand) < 0) &&
370 (o_hand.sa_handler == SIG_IGN) &&
371 (sigaction(SIGINT, &o_hand, &o_hand) < 0))
372 goto out;
373
374 if ((sigaction(SIGQUIT, &n_hand, &o_hand) < 0) &&
375 (o_hand.sa_handler == SIG_IGN) &&
376 (sigaction(SIGQUIT, &o_hand, &o_hand) < 0))
377 goto out;
378
379 if ((sigaction(SIGXCPU, &n_hand, &o_hand) < 0) &&
380 (o_hand.sa_handler == SIG_IGN) &&
381 (sigaction(SIGXCPU, &o_hand, &o_hand) < 0))
382 goto out;
383
384 n_hand.sa_handler = SIG_IGN;
385 if ((sigaction(SIGPIPE, &n_hand, &o_hand) < 0) ||
386 (sigaction(SIGXFSZ, &n_hand, &o_hand) < 0))
387 goto out;
388 return(0);
389
390 out:
391 syswarn(1, errno, "Unable to set up signal handler");
392 return(-1);
393 }
394